ASML Holding, the Dutch giant with a near-monopoly on extreme ultraviolet (EUV) lithography machines essential for advanced AI chips, is well-positioned to capitalize on the data center surge. Yet, its stock has risen only 11% in the past year, underperforming peers, due to warnings of uncertain 2026 growth. The core issue: reliance on a handful of major customers, dominated by Taiwan Semiconductor Manufacturing (TSMC), which controls advanced chip production. ASML thrives on competitive battles among chipmakers that spur equipment upgrades, but TSMC's lead stifles that dynamic.
ASML's latest innovation, high numerical aperture (High NA) EUV machines—costing over $400 million each—promises single-patterning for smaller, more efficient chips, reducing complexity, cycle time, and defects compared to multi-patterning on older EUVs. However, TSMC is hesitant, citing high upfront costs and plans to extend current tools' lifespan until High NA matures for maximum value. Barclays analyst Simon Coles forecasts just three High NA shipments in 2026 (down from five in 2025), with broad adoption delayed until 2028; he rates ASML a Hold.
Intel emerges as ASML's potential savior, seeking redemption after lagging TSMC by delaying EUV adoption. Former CEO Pat Gelsinger committed to two High NA machines to challenge TSMC in AI chip foundry services via Intel's 18A process. Yet, Intel faces hurdles: mastering EUV yields, ecosystem readiness, cost competitiveness, and securing external clients like Nvidia and Broadcom. Recent hype around Nvidia's $5 billion investment in Intel boosted ASML shares 6%, but Nvidia's CEO Jensen Huang dodged questions on using Intel fabs, praising TSMC's superiority for complex GPUs. Analysts like Nicolas Baratte question Intel's need for pricier High NA if TSMC deems it unnecessary. Intel has two machines in Oregon R&D but won't confirm production timelines, and it may scrap its 14A process without major customers, jeopardizing U.S. semiconductor ambitions.
Memory chips offer another avenue. Lagging logic in EUV adoption due to stacking techniques and cost focus, the sector is shifting with high-bandwidth memory (HBM) demand for Nvidia's AI processors. SK Hynix leads, assembling the first High NA system for mass production to meet scaling needs. Rival Samsung adopted High NA in February for memory and foundry work, pressuring U.S.-based Micron to follow.
Ultimately, ASML's superior tech can't guarantee quick wins amid customer dominance. High NA will become indispensable, but investors may endure delays as the industry weighs costs against future gains. (Word count: 428)